Category: COVID Archives

National growth rate drops to 3.3% – GA, MI, PA added

The national growth rate dropped from 3.8% to 3.3% yesterday, which is again the lowest growth rate we’ve seen to date. However, I was expecting a larger fall. Let’s hope for that today. I’ve added a daily report on GA, MI, and PA. All three states are past peak active cases, and in decline. In addition, the number of reported pending tests is very small.

IHME published a revision to their COVID model late yesterday. They have added an estimate for each state as to when they believe lifting restrictions is prudent. Their criteria, however, are when infections drop below 1 per million of population, or 0.0001%. That would mean less than 11 cases in NC, for example. I don’t believe we will ever get to this threshold. In reality, states are beginning to move more quickly than this. More below.

As always, if you have any questions about my assumptions, or how I’m modeling, or want to know about Logistic modeling generally, drop me a line.

• Likely date of active case peak (Chalke modeling): April 10
• Likely date of peak deaths (IHME): April 15 (last revision on April 21)
• Short term projection for tonight: 852,000
• Total Test Results reported yesterday: 151,627
• Total Pending tests reported yesterday: 3,956 (a new low)
• National reported case Growth Rate yesterday: 3.3% (a new low)

National growth rate drops to 3.8% – South Carolina added

The national growth rate dropped from 4.0% to 3.8% yesterday, which is the lowest growth rate we’ve seen to date. I’m expecting the growth rate to continue to fall the remainder of this month. It looks like the first 3 states to begin opening back up are SC, TN, and GA. I’ll add all three to the daily report as I have time, since I’m very curious to see what, if anything, changes when they reduce restrictions. I’ve added South Carolina today.

My Logistic model projected 800,564 cases yesterday, and the actual numbers came in at 800,964, or a deviation of 0.0%. That’s no big deal, as the course of this disease has been pretty easy to predict in the short run. The test will be how the model holds up over the next 4 or 5 days, as the model projects a declining new case count.

As always, if you have any questions about my assumptions, or how I’m modeling, or want to know about Logistic modeling generally, drop me a line.

• Likely date of active case peak (Chalke modeling): April 10
• Likely date of peak deaths (IHME): April 15 (last revision on April 17)
• Short term projection for tonight: 823,000
• Total Test Results reported yesterday: 137,687
• Total Pending tests reported yesterday: 4,037 (a new low)
• National reported case Growth Rate yesterday: 3.8% (a new low)

National growth rate 4.0% – possible peak in Massachusetts

Pretty good news again today. The national growth rate stayed at 4.0%, which is the lowest point we’ve seen to date. We usually see a drop in reporting on Sundays, but not this week. We saw the 2nd highest reported test day, with over 175,000 test results reported, so that gives me confidence that Sunday was not an aberration.

I refit my Logistic model on Saturday afternoon, and I’ve received several questions about what that means. Early on, I chose a functional form that is the one most often used for modeling disease propagation. I have recalibrated the model twice now to historical data, but haven’t changed the model form, which looks like this:

A picture containing text, whiteboard, bird, flying

Description automatically generated

If that’s TMI, I apologize. In any event, models remain effective by learning from changes in the data over time. There are many reasons to recalibrate models, but in this case we’ve seen a lot of moving parts. Specifically, the rapid change in scope of testing, constantly changing definition of what a “reported case” is, and some historical reclassification of data. So we’ll use this model so long as it reasonably tracks actual reported data from day to day, and when something changes where the model is no longer useful, I’ll recalibrate it again. It looks like IHME revises their model about every three days. I’ll likely let mine ride much longer than that. Before active cases peaked, I used the model primarily to forecast peak cases, and it was quite close. Now I’m using the model to project when we’ll see active cases significantly decline.

As always, if you have any questions about my assumptions, or how I’m modeling, or want to know about Logistic modeling generally, drop me a line.

• Likely date of active case peak (Chalke modeling): April 10
• Likely date of peak deaths (IHME): April 15 (last revision on April 17)
• Short term projection for tonight: 800,000
• Total Test Results reported yesterday: 175,382
• Total Pending tests reported yesterday: 11,324
• National reported case Growth Rate yesterday: 4.0% (low point to date)

National growth rate falls to new low of 4.0%

The news today is better. The national growth rate fell from 5.2% to 4.0%. We witnessed a small reversal in the trend of the data the past few days. Some of it was the reclassification of data by NYC, but it was more widespread than that. As of April 14, the CDC changed the definition of a “case”. Here’s an excerpt from the CDC website:

2. As of April 14, 2020, CDC case counts and death counts include both confirmed and probable cases and deaths. This change was made to reflect an interim COVID-19 position statementpdf iconexternal icon issued by the Council for State and Territorial Epidemiologists on April 5, 2020. The position statement included a case definition and made COVID-19 a nationally notifiable disease.

A confirmed case or death is defined by meeting confirmatory laboratory evidence for COVID-19. A probable case or death is defined by i) meeting clinical criteria AND epidemiologic evidence with no confirmatory laboratory testing performed for COVID-19; or ii) meeting presumptive laboratory evidence AND either clinical criteria OR epidemiologic evidence; or iii) meeting vital records criteria with no confirmatory laboratory testing performed for COVID19.

You’ve probably seen the headlines about the spike in COVID deaths, but this is caused by data redefinition, and doesn’t mean that the disease is suddenly accelerating. As we saw a couple of weeks ago when the case definition was last expanded, the case count grew at a faster rate for a few days, then settled down to the previous pattern. I believe we’ll see the same thing this time. This does make modeling more difficult without consistent case definitions. I’ve experienced this in the past during the AIDS epidemic. There was a constant widening of the definition of an AIDS case, and part of our job was to figure out how much was actual growth and how much was reclassification. I’m not even attempting that here. I’m just letting the data flow through, and doing my best to explain it.

In other news, I did re-parameterize my Logistic model yesterday. This fit is not as tight as the last fitting, and my theory is that in fact, the progression has slowed over the past couple of weeks as a result of societal behavioral changes. This is probably of more interest to me than to you, but I’ll use the newly refitted model in the graphs where I’ve contrasted to model numbers.

As always, if you have any questions about my assumptions, or how I’m modeling, or want to know about Logistic modeling generally, drop me a line.

• Likely date of active case peak (Chalke modeling): April 10
• Likely date of peak deaths (IHME): April 15 (last revision on April 17)
• Short term projection for tonight: 770,000
• Total Test Results reported yesterday: 149,114
• Total Pending tests reported yesterday: 9,906 (new low)
• National reported case Growth Rate yesterday: 4.0% (new low)

Massachusetts added – IHME Model update

The national growth rate rose again yesterday, from 5.0% to 5.2%. I spend hours every day looking for patterns in the data, so have become quite familiar with the ebb and flo of the data each day. Very little reporting happens before noon, and we get significant reporting around 1pm, again around 4:30, and once again around 8pm. Little happens after 9pm, so I start looking at the day’s reports around then. Last night things were looking very good, with a nice drop in new cases, and the national growth rate dropping to 4.4%. For some reason I went to bed early last night, around 11pm. When I wake up this morning about 5,000 new cases were added to yesterday’s data. I don’t know what to make of this, but as a mathematician, I like regular patterns. I’m going to watch this as it emerges tonight to see if this repeats, and try and determine where the late reporting is coming from.

The IHME Model released an update yesterday, and the news is good. They now project that peak deaths happened on April 15, and have lowered their total death projection to 60,000. They are now showing peak deaths in NY on April 16. The biggest shock is Florida, where they previously projected peak deaths on May 9. They are now projecting that peak deaths already occurred some days ago, on April 2. Since I calculate peak cases in Florida on April 8, I would have expected peak deaths to be more like April 15. We’ll see.

I have added Massachusetts to my analysis, as the individual states I’ve been tracking are all past or around their active case peak. Massachusetts, however, is still experiencing active case growth, so it is worthwhile to track.

Finally, I’m going to refit my Logistic Model this weekend, and will have a better overall projection of the decline.

As always, if you have any questions about my assumptions, or how I’m modeling, or want to know about Logistic modeling generally, drop me a line.

• Likely date of active case peak (Chalke modeling): April 10
• Likely date of peak deaths (IHME): April 15 (last revision on April 17)
• Short term projection for tonight: 749,000
• Total Test Results reported yesterday: 140,304
• Total Pending tests reported yesterday: 10,889 (a new low)
• National reported case Growth Rate yesterday: 5.2%

Small uptick in growth rate today

An abbreviated report today as I’m tight on time. More tomorrow. The national growth rate rose slightly yesterday from 4.9% to 5.0%. I want to spend some time over the weekend to look at all the states growth rates to see where this is coming from, but it’s not NY data this time. NY reported a decrease in new cases yesterday, from 11,571 to 8,505. Of course, the previous day’s number (11,571) was a marked increase from the day before, and I still don’t know whether the reclassified 3,778 deaths were included in Wednesday’s reporting. I do not try and correct for changing definitions or protocol. I do consistently use the data from just 3 sources: www.infection2020.com for national daily numbers, the COVID Tracking Project for daily state data, and the IHME model for any comments about peak deaths. I’m anxiously awaiting an update to the IHME model, which was last updated on April 13.

As always, if you have any questions about my assumptions, or how I’m modeling, or want to know about Logistic modeling generally, drop me a line.

• Likely days past active case peak (Chalke modeling): 7 days
• Likely days past peak deaths (IHME): 4 days (last revision on April 13)
• Short term projection for tonight: 710,000
• Total Test Results reported yesterday: 158,309
• Total Pending tests reported yesterday: 16,927
• National reported case Growth Rate yesterday: 5.0%

NYC adds a wrinkle

The national growth rate rose yesterday to 4.9%. This increase is almost entirely due to NY data, and I’m still trying to make sense of the historical revision, and which numbers it has impacted and which is has not. You may have read that on Tuesday NYC reclassified 3,778 deaths going back to March 11 as COVID deaths. I have no issue with this, but I’m trying to accurately determine when and how these new cases are reported in the data. It appears that the state of NY has revised the death count numbers historically. As a result, there is no single day where you see the additional 3,778 deaths reported. On the other hand, it appears that the case count history was not revised, and I believe (but don’t know) that the reclassified deaths were reported as new cases yesterday. NY reported 11,571 new cases yesterday, in contrast to the previous 4 day average of 7,924, or a bump of about 3,647. If any of you have any first hand knowledge of NY’s reporting methodology for the revision, please drop me a line.

In any event, we’ll have a better idea later today if we see new case reporting in NY decline again. I’m not going to try and correct for the revision – I’ll just let it pass through the numbers. It’s likely a one-time perturbation of the data. Even if other localities follow suit (some will), the numbers are all low compared to NY, where the mass of the disease is centered.

According to my modeling, and even with the change in NY data, I believe we’re down to about 210,000 active case in the U.S. More below.

As always, if you have any questions about my assumptions, or how I’m modeling, or want to know about Logistic modeling generally, drop me a line.

• Likely days past active case peak (Chalke modeling): 6 days
• Likely days past peak deaths (IHME): 3 days
• Short term projection for tonight: 676,000
• Total Test Results reported yesterday: 161,135
• Total Pending tests reported yesterday: 16,901
• National reported case Growth Rate yesterday: 4.9%

National growth rate hits new low of 5.2%

As I said yesterday, I don’t totally trust Easter Weekend data, but nonetheless the news is good. We hit a new low national growth rate yesterday of 5.2%. The IHME model shows that we’re 3 days past peak deaths in the United States, and my model shows 3 days past peak active cases. One place where my model differs from the IHME is in Florida. I show Florida active cases peaking a few days ago, but IHME doesn’t project peak deaths in Florida until April 27. That’s two weeks away. I would have expected peak deaths earlier. In any event, IHME projects total COVID deaths in Florida at less than 4,000 through mid-summer. That is just a tiny fraction of total annual deaths in Florida from all causes. In NC, we’ve been at or just around the peak for days, but we haven’t seen the steady decline yet like we see in Florida or California. Of course, there is very little COVID in NC, so the sampling isn’t as large. In New York, where they have suffered the most, we are now well past peak active cases, and peak deaths.

I was expecting a flood of reporting today after the holiday weekend, but as I write this it’s 1:09pm and very little so far. I am expecting a small uptick today, but will be happy if we don’t see it.

Note that the IHME model shows national peak deaths 3 days ago, and my model shows national peak cases also 3 days ago. These two data points are somewhat incongruous, as Peak Cases precedes Peak Deaths – I would have though by a week or so. We know that we are in fact past peak deaths, because deaths are reported. We’re less certain about the day we reached peak active cases, because recoveries are not reported in any meaningful way. It’s likely that we reached peak active cases earlier than I’m showing.

As always, if you have any questions about my assumptions, or how I’m modeling, or even want to know the functional form of the model, drop me a line.

• Likely days past active case peak (Chalke modeling): 3 days
• Likely days past peak deaths (IHME): 3 days
• Short term projection for tonight: 592,000
• Total Test Results reported yesterday: 139,870
• Total Pending tests reported yesterday: 16,419
• National reported case Growth Rate yesterday: 5.2%

The worst is over

Happy Easter! I’m feeling optimistic today. The national growth rate fell to 5.6%, and it looks like the U.S. peaked at something less than 250,000 active cases and is on the way down. Of course, the news is slow to react to this, but by this time next week all anyone will be talking about is the “recovery”. Of course, please send me any questions you may have on the data or my analysis.

• Current Prediction of Reported and Still Active Case Peak: April 11
• Confidence: Very high
• Short term projection for tonight: 563,000
• Total Test Results reported yesterday: 137,297
• National reported case Growth Rate yesterday: 5.6%

The news focuses on total cumulative reported cases. This is an important number, but not the most important. Far more critical is the number of people who are currently infected. We know, for example, that someone infected with the disease in early March is now recovered, or (far, far less likely) deceased. Yet they are still counted in the headlines. When you see a headline that says “US passes 500,000 cases”, this includes (by my calculations) over 250,000 people that have recovered. We must make assumptions about this number because recoveries are generally not reported. I’ve made the best assumptions I can about time from disease onset to recovery, and I think they are sound. Apologies if this little discussion insults your intelligence, but this list is quite diverse in people’s comfort with data or math. On this list are Mathematicians, Scientists, Actuaries, Analysts, Engineers, Artists, Dancers, and Musicians (a sampling of my life). In fact, about ½ dozen of you are both mathematicians AND musicians. And about ½ the people on this list I don’t even know. So I’ll mix up a little math with a little explanation each day.

Here is the graph of model vs. actual active reported cases. The model turned out to be about 3 days ahead of the reported results, but the shape will be the same. After all the hysterical news of the past 2 weeks, this is a beautiful thing to see. One thing to note is that this IS Easter weekend, so I expect some irregularities in the data, which I expect to smooth out by Tuesday. However, I think yesterday’s results were pretty solid, and we’ve not seen any pattern of incongruous results on a Saturday yet. Yesterday 137, 297 test results were reported, compared to a 5 day average of 149,785. Also, the testing backlog is still very low at 16.5K. In fact, one of the doctors on this list reported to me that his hospital now processes test results in one hour, so I expect backlogs to remain insignificant. The net result of all this is that I’m not entirely trusting the data until Tuesday. If we get any extraordinary reporting, I expect it’s today. In any event, stare at the graph below and enjoy it. This result is also corroborated by the IHME model, which reports that peak deaths in the U.S. happened two day ago, on April 10th.

Looks like we’re over the hill nationally, CA peaks

Please feel free to send me any questions you may have on the data or my analysis – I welcome all thoughts. Or just drop me a line and let me know if you find this daily report useful.

• Current Prediction of Reported and Still Active Case Peak: April 10 or 11
• Confidence: High
• Short term projection for tonight: 542,000
• Key Stats:
• Total Test Results reported yesterday: 168,213
• National reported case Growth Rate yesterday: 7.6%

So all the headlines today are focusing on the fact that the U.S. now has over 500,000 cumulative reported cases. What they’re not saying is the progression of the disease has slowed to a crawl, and we’re at the point of maximum active cases. I know it’s hard to see just looking at a graph of cumulative cases. It’s probably obvious, but cumulative reported cases cannot ever decrease. Once as case is reported, it stays reported. However, this new headline does not mean 500,000 people have COVID. It means that 500,000 people have or have had COVID. When we talk about Peak Active Cases, we refer to the maximum number of people at any given time that have the disease. Just now it looks like something less than 250,000.

The national growth rate fell yesterday from 7.9% to 7.6%. It’s clear that active cases are peaking. It’s the picture below that’s important. We’ve been waiting for the blue line to level off. The peak was yesterday, or perhaps today, but they won’t differ by more than a couple thousand cases. The United States is peaking at something less than 250,000 active cases, which is good news, relative to everything you’ve read over the past couple of weeks. According to the IHME model, peak deaths in the U.S. occurred yesterday. Unless the nature of the disease changes, peak cases must occur before peak deaths, so perhaps the lag time between case reporting and death reporting is shorter than I’m assuming.